Build Your First AI Project This Weekend
Stop consuming tutorials. Start creating. Get the free step-by-step guide.
2026 Update:
Outbrain the company is gone as a name. Outbrain bought Teads (deal closed February 3, 2025) and renamed the parent Teads Holding Co. on June 9, 2025. Nasdaq ticker is TEAD (GlobeNewswire). You will still see “Recommended by Outbrain” widgets on publisher pages. The advertiser relationship is Teads.
Taboola is still the big direct-response native pipe. Exclusive Yahoo native for 30 years, announced as covering hundreds of millions of Yahoo users a month, on top of 9,000+ publisher partners (Taboola). Their advertiser UI is now Realize. In 2026 they have been pushing past the chumbox into more display on some publishers (NBC News programmatic display is one example).
AI is useful here because native is a headline factory. It is not a substitute for an advertorial that matches the click.
Who this is for: affiliates, info products, content sites, lead magnets that can be told as a story.
Who this is not for: a dentist who wants new patients this week. That is Google Ads and local SEO. Nobody reads a Yahoo article and books a root canal from a “10 secrets dentists don’t want you to know” widget unless you like refunds and chargebacks.
What native advertising is
The “You may also like” / “Around the web” tiles under a news article. Thumbnail + headline. Paid. Looks like another article.
The user is in read-mode, not search-mode. You win with curiosity and a page that feels like content. You lose with “50% off today” banner energy.
That is why native loves advertorials and hates naked squeeze pages.
Networks in 2026
Taboola (Realize)
Scale and premium pubs. Yahoo, lots of news, weather, OEMs.
Pros: volume, better brand-safe inventory, mature bidding and creative tools.
Cons: spend. Managed accounts want real monthly budgets. Review is picky. CPC is not pop-cheap.
Use when: the funnel already converts somewhere and you can feed the machine 10+ headlines.
Teads (was Outbrain)
Premium pubs, more video/CTV in the pitch than “native CPC arbitrage.” Performance buyers still run recommendation inventory. Expect branding language on sales calls.
Pros: CNN-class placements historically, conversion bidding modes, brand safety.
Cons: learning phase, tighter creative rules, higher CPC.
Use when: you graduated from MGID and can pass conversion data.
MGID
Where a solo operator should test first. Lower minimums, sloppier inventory, more work on the publisher report.
Revcontent
Between MGID and the big two. Still needs blacklists.
I am not sending you to every 2017 “native alternative” on a listicle. If you cannot get approved at MGID, native may not want your offer.
How the buy works
Usually CPC. Bid more, see better tiles. Quality score / CTR feeds volume at the same bid. A 2% CTR headline will bury a 0.2% headline.
Some platforms offer target CPA / max conversions once you have conversion volume. Do not turn those on with 12 clicks of data.
Start CPC. Add conversion optimization after the pixel is honest.
Headlines and images (the whole sport)
Taboola’s public creative advice has long been: roughly 34-45 characters, 60 max, front-load the point. I still write to that even on MGID because mobile truncates everything.
Formulas that get clicks without being a lie:
- Curiosity gap that the page actually fills
- Numbered list (“7 mistakes…”)
- Specific claim you can prove
- Named audience (“freelancers over 40”)
- How-to
- Mild warning (networks reject gore and fake disease)
Avoid: giving away the answer, !!!, celebrity lookalikes, “you won’t believe,” medical miracles, income claims without a landing page that can survive review.
Images: 800x600 or larger. Real faces, tight crop, editorial feel. No six lines of WordArt. No before/after bodies if the network bans them (most do).
Minimum test: 5 headlines x 2 images. Better: 10 x 5. Most combos fail. Two will carry spend.
Headline prompt
You write native ads.
Offer: [X]
Audience: [Y]
Benefit: [Z]
Angle: [curiosity | how-to | warning | authority]
20 headlines, 34-45 characters, 60 max.
Editorial, not "buy now."
No fake news masthead, no celebrity.
1. [headline] - [count] - [formula]
Feed winners back in: “Here are 5 titles with CTR. Make 20 cousins.”
Image concepts
Native images for [offer].
Headlines: [3-5]
10 concepts: subject, emotion, close-up vs wide, why it stops a scroller.
Editorial, not stock-handshake.
Midjourney / stock / a $50 Fiverr batch. You do not need a studio.
Advertorial or go home
Native traffic bounces on a thin landing page. Structure:
- Headline that pays off the ad
- Hook
- Problem story
- Discovery of the product
- Proof
- Benefits
- CTAs in the body, not one button at pixel 4000
Must-haves for review: sponsored disclosure, real contact, privacy policy, no fake comments, no fake NYT byline.
Full build: landing pages. Tracking: tracking guide.
Advertorial prompt:
Advertorial for [offer].
Ad headline that got the click: [headline]
Audience, problem, 3-5 benefits, proof I actually have: [list]
1,500 words, conversational, multiple CTAs, not a fake newspaper.
Edit it. AI first drafts read like AI. Reviewers can tell. So can users.
Bidding and the publisher report
Rough CPC bands (your vertical will disagree):
- Taboola / Teads: start at their suggested bid, often tens of cents to $1+
- Revcontent: mid
- MGID: cheaper, noisier
On conversion bidding, daily budget needs to be a multiple of target CPA or the algorithm starves. Do not smash budget every day. Do not layer manual CPC on top of target CPA.
The real work is publishers.
- Track site / widget
- Blacklist sinks
- Whitelist converters
- Eventually run a whitelist-only campaign at a higher bid
This is tedious. This is the profit. AI can rank the CSV. You still click block.
Publisher report.
Volume threshold: [N] clicks.
Top 10 by CR, bottom 10 to block, patterns, what needs more data.
[paste]
Daypart if the data is obvious. Do not daypart on 40 conversions.
CTR sanity: many native campaigns live between a fraction of a percent and about 1%. A “2% CTR” is a star, not a baseline. Kill true zeros. Do not kill 0.3% on day one.
What to sell (and what not to)
Fits: productized services with a lead magnet, courses with a quiz or webinar, software trials, affiliate offers that can be a story (finance, education, consumer software).
Does not fit: niche B2B with a 200-person TAM, $5k with no step in between, local services, pure impulse “buy this toaster.”
If you need someone who typed a keyword, buy search.
First month, no theater
Prereqs: an offer that has converted somehow, $500-1,000 you can lose, tracking live.
Week 1: 50 headlines, 10 images, advertorial, MGID or Revcontent, $50/day, US, leave it alone.
Week 2: watch, do not poke.
Week 3: block bottom publishers, kill dead titles, clone winners.
Week 4: if CPA is in range, raise budget 20-30% and think about Teads/Taboola. If not, fix offer or page before you “scale.”
DCO (dynamic combinations of your headlines and images) is useful once you uploaded enough parts. It is not magic with two titles.
I am not going to quote “30-50% CTR lift” from a vendor PDF. Upload volume, then read your own CTR.
Conversion bidding without folklore
When the platform offers target CPA / max conversions:
- Pixel or postback must be firing on the real event, not “page view”
- Daily budget should be several times the target CPA or the system cannot buy a conversion
- Leave it alone for days, not hours
- Do not also sit there pulling publisher blocks every morning during the learning window (you can still kill obvious fraud domains)
- If you need a hard CPC cap, you may not want fully automatic mode yet
I start manual CPC on MGID. I turn on conversion bidding on Taboola/Teads after the advertorial already converts on cheaper inventory.
Creative refresh
Native tiles live next to real headlines. Users go numb. Plan:
- Weekly: new titles that cousin the winners
- Monthly: new images, not just a recolor
- When CTR on a winner drops for 7+ days with spend still there, it is tired. Do not raise bids on a tired tile.
AI’s job is the cousins. Your job is killing the original when it is done.
Compliance that gets accounts banned
Fake news chrome (CNN logo, “breaking”). Fake doctors. Income screenshots you cannot back up. Before/after bodies. “You are the 1,000,000th visitor.” Comment widgets with stock people.
If you need that to convert, native is the wrong channel and the offer is the wrong offer.
Mistakes
- One title, one image
- Ignoring publishers
- A sales page wearing a fake article hat
- Scaling 5x in a day
- Native for a plumber
- Uploading two headlines and calling it DCO
- Changing budget every day, then blaming the algorithm
A note on “Yahoo traffic”
Yahoo via Taboola is premium-ish inventory, not a magic geo. You still need the advertorial. You still need to pass review. You will still see widgets on sites you do not want. The 30-year deal is why Taboola has scale (Taboola). It is not why your CPA will be $3.
If a sales rep pitches “Yahoo exclusive” as a substitute for a funnel, smile and send them your landing page.
Should you use native?
Yes: story-shaped offer, weekly optimization time, stomach for a learning bill, you want off Meta/Google.
No: high-intent search is the job, no time, no loss tolerance, ultra-niche, set-and-forget fantasy.
Most people quit native in week two because they expected Facebook-style audience sliders. You do not pick “homeowners 35-54 who like golf.” You pick a story, a bid, and a publisher list. That is the job.
Next
Have a funnel: MGID account, headline prompt, $50/day, seven days hands-off.
No funnel: Plan first.
Local service: Google Ads and local SEO.
Related: paid traffic map | RTB | push | AI tools for media buyers
Questions? Contact me.